Last week, Goldman Sachs doubled its target price for Zhongji Innolight, a Chinese optical module manufacturer that most crypto natives have never heard of. The rationale? Silicon photonics shipments are ramping up, and the “Scale-up” networking market is exploding alongside AI server racks. As a crypto educator who has spent years dissecting the layers of blockchain infrastructure, this news hit me differently.
We’ve been obsessed with software decentralization—smart contracts, DAO governance, L2 rollups. But beneath that software lies a physical layer of hardware. And that hardware layer is becoming terrifyingly centralized. The same bottlenecks that limit AI model training—network bandwidth and latency—are starting to choke blockchain scalability. Unlike AI, however, crypto’s ethos demands resilience through distribution. A single optical module supplier becoming the backbone of the next generation of compute clusters should give every decentralization advocate pause.
Context
Let’s unpack what Zhongji Innolight actually does. They manufacture high-speed optical transceivers that convert electrical signals into light signals and back. These modules are the arteries of modern data centers, connecting thousands of GPUs or ASICs within a cluster. Silicon photonics—a technology that integrates optical components onto silicon chips using standard CMOS processes—is their secret sauce. It promises lower cost, lower power, and higher integration density compared to legacy III-V semiconductor approaches. Goldman’s analysts specifically cited growing shipments of silicon photonics-based modules as a key driver. They also highlighted the shift from “Scale-out” networks (connecting many servers loosely) to “Scale-up” networks (high-bandwidth connections within a compute node, like NVIDIA’s NVLink). This shift is exactly what enables monster AI clusters like the DGX GB200 NVL72.
But here’s where the crypto connection sharpens. Bitcoin mining relies on ASICs connected to pools via the internet; Ethereum validators run on cloud instances with varying network specs; decentralized storage networks like Filecoin need high-throughput links between nodes. While we obsess over consensus algorithms and tokenomics, the physical network that binds these nodes together is often an afterthought. The harsh reality is that the global supply of high-speed optical modules is controlled by a handful of companies: Zhongji Innolight and Coherent dominate the 800G era. This is a single point of failure waiting to be exploited—or regulated.
Core
The technical analysis here cuts both ways. On one hand, better optical modules mean blockchain nodes can sync faster, handle larger blocks, and reduce latency. For networks like Solana or Aptos that aim for thousands of transactions per second, high-bandwidth interconnects are not a luxury; they are a necessity. The shift to Scale-up networking in AI mirrors the trend in crypto toward monolithic blockchains where all validators process the same transactions in parallel. The demand for 1.6T optical modules will inevitably spill over into the crypto infrastructure market.
On the other hand, this creates a stark centralization pressure. My own journey through the 2020 DeFi summer destruction taught me this lesson painfully. After I lost $15,000 in a yield farming exploit, I spent months reverse-engineering the hack. It turned out the root cause wasn’t a smart contract bug but a front-running bot that exploited oracle latency mediated by a single fast node. The hardware pipe was the weakest link. That experience lodged in my mind: the cost of high-speed networking becomes a moat that only well-funded entities can cross. If the next generation of blockchain nodes demands 1.6T optical modules that cost tens of thousands of dollars each, independent operators will be priced out. Node counts will drop, and the network will grow more vulnerable to capture.
Consider the parallel to L2 sequencers. For two years, we’ve heard promises of “decentralized sequencing,” yet most L2s still run on a single sequencer operated by the development team. The sequencer is essentially a centralized node that batches transactions and posts them to the base layer. The excuse is always the same: “We need performance first; decentralization comes later.” Zhongji Innolight’s rise signals that the hardware barrier to decentralization is only getting higher. The same pattern emerges in Bitcoin mining: concentrated among pools and ASIC manufacturers. We have not solved the hardware centralization problem; we have only anesthetized ourselves with software narratives.
Contrarian
Now comes the counter-intuitive angle: maybe this centralization is not just inevitable but optimal for performance. The market signal from Goldman is capital flowing to the most efficient provider. Perhaps crypto should accept that some hardware layers must be centralized to achieve the throughput needed for mainstream adoption.
But truth in blockchain isn’t about acceptance—it’s about resistance. We didn’t build Bitcoin to replicate the power structures of centralized finance. We built it to provide an alternative even if it is slower. The same principle applies to hardware. The contrarian stance here is that the crypto industry should actively fund and research alternative network topologies and modular hardware designs that reduce reliance on a single optical module supplier. Projects like Helium’s decentralized wireless network or Filecoin’s proof-of-replication already experiment with commodity hardware. But they are exceptions. The broader infrastructure for high-speed blockchain consensus remains dangerously dependent on the same few companies that serve AI.
I’ll be honest: Goldman’s upgrade makes perfect financial sense. Zhongji Innolight is riding a wave that will make it even more critical. But for those of us who hold decentralization as a core value, this should be a wake-up call, not a celebration. The next crypto bull run may not be about DeFi or NFTs—it may be about who controls the pipes. We didn’t fight to replace banks with centralized sequencers and single-supplier optics.
Takeaway
The future of crypto depends on whether we can build open, distributed hardware supply chains. It’s time for DePIN projects to step up and prove that resilient infrastructure is possible without sacrificing performance. Otherwise, Goldman’s target price will become our technical reality—and that reality is not decentralized.
